Safran Aircraft Engines — Industrial Division · 2025
Predicting the energy consumption of an industrial machine
Predictive modelling of a production asset's consumption to uncover energy-saving levers.
potential annual savings identified
explainable predictive model
critical asset modelled
The challenge
The machine's energy consumption was only reviewed after the fact, with no way to link it to operating conditions or anticipate drifts.
The approach
- 1Collected and cleaned operating and consumption history.
- 2Exploratory analysis to identify explanatory variables.
- 3Trained and compared regression models with scikit-learn.
- 4Translated results into concrete levers for the teams.
Solution architecture
Data
Operating and consumption history
Preparation
Cleaning and feature engineering (pandas, NumPy)
Modelling
scikit-learn regression models, cross-validation
Delivery
Savings scenarios and recommendations
The results
- Potential savings of about 12% on annual consumption identified.
- A detailed understanding of what drives consumption.
- A reusable foundation to extend the approach to other assets.
What this project shows
“Machine learning creates value when it leads to a decision: here, concrete and quantified savings levers.”
These projects were led by Datasium's founder during assignments and internships before the firm was founded. Figures are those observed at the end of each assignment.